Translation memory’s a timeworn phrase in the translation industry. But what actually is it? And, more importantly, how can translation memory save money on translation?

So what exactly is translation memory?

Translation memory (TM) is a linguistic database that stores previously translated text. It’s used alongside computer-aided translation (CAT) tools like TRADOS and DejàVu.

A translation memory is basically a continually evolving digital library that contains all of the translations that have been undertaken for your company. These are stored as segments. Segments are translated sentences, phrases, headings and other sections of text.

When a translator’s working on a document, any segments that are repeated from a previous translation are matched and extracted from the TM.

Any exact matches as well as any partial (fuzzy) matches are placed within the document for the translator to review. The translator then works through the document translating the non-repeated text, while reviewing any matches as they come across them. They can choose to accept the matches or override them with a new alternative.

But how does this differ from cheap machine translation?

If you’re new to the world of translation, you might worry that computer-aided translation is just another form of cheap machine translation. But it’s not.

Machine translation results in a low-quality translation, since it automatically translates text without any human intervention. Subtleties of language like tone, context and colloquialisms are often missed out. Machines translate directly word-for-word. Inferred and implied meanings are always overlooked. We’ve written about this in detail here.

Translation memory doesn’t result in a low-quality translation. By using translation memory, translations are more consistent and of a higher quality.

OK. So how do translation memories improve translation consistency?

Most text for translation has repeated or similar content. And when this content needs to be translated again and again throughout a document or website, it can be near impossible – and laborious – for a translator alone to ensure consistency. They would need to keep scrolling back through the text to check their previous translations. If they use a translation memory, this step is done for them.

So by working with a translation memory things like terminology, tone and style are kept consistent throughout the translation, as they are always translated in the same way.

And this is the case even if it is another translator working on a document at another time, or if a team of translators are working on a high-volume document at the same time.

When should a translation memory be used?

In principle translation memories can be used to aid any type of translation.

In practice, translation memories work best with material that contains a lot of repeated or partially-repeated sentences and phrases.

Because of this they are particularly useful when translating:
• Instruction manuals
• Large websites
• Software documentation, help files and FAQs
• Regularly repeated content – think newsletters and periodicals, monthly reports, regularly updated web content
• Large word-count documents that are likely to have repeated content

Translation memory is also useful if you have a regular need for translation. It might be that the documents themselves don’t have a lot of repetition, but translation memory will aid consistency in tone and style across multiple translated documents.

So how does translation memory save money on translation?

Translation costs can be discounted when there is repeated content. And translation memories analyze where repeated content appears in a document. The per-word rate is reduced since the translator doesn’t need to translate content that is stored in a TM.

Before translation begins, the document will be analyzed against the translation memory. This shows the number of exact matches and fuzzy matches in the document compared to the translation memory.

This is then broken down into categories, and a proportionate discount is applied. A higher discount will be applied to exact matches, since these won’t require translation. Fuzzy matches may be discounted, but translators will still need to review and edit these, so the discount will generally be less.

Are there any other advantages to using a translation memory?

Alongside an increase in consistency and a reduction in translation costs, translation memory also speeds up the translation process.

The same translation segment never needs translating twice. This significantly reduces the time it takes a linguist to translate new material, so more time can be spent on reviewing the translation for grammatical and contextual accuracy and consistency.

Plus, by speeding up the translation process, it becomes much easier for a translator to stick to a tight deadline, and to turnaround a project quickly.

Without using a translation memory to capture repeated content, your translator would have to manually translate the same phrases, terminology and sentences again and again: a time-consuming and costly task for the translator and, in turn, for you.

Do you need help with translation?

Get a quote View our case studies


Sure Languages is a professional translation company based in the UK. We help businesses from all sectors communicate in over 100 languages through our specialised translation, interpreting and voiceover services.